WO2012046671A1 - Positioning system - Google Patents

Abstract

In a positioning system for identifying the current position of a mobile body by comparing a landscape image captured by image capture means provided in a mobile body with a landscape image database in which features of landscapes have been described, a landscape image database of the positioning system is provided so as to perform positioning processing in the mobile body to be positioned using less calculation. The landscape image database of the present invention is such that a plurality of landscape image data and image acquisition positions, which are positions where the landscape image data were gathered, are associated and stored, wherein each of the plurality of landscape image data includes a feature value of a feature point corresponding to an entity other than entities for which the possibility is low that the position and shape in the real world will be maintained as-is for a predetermined period or longer.

Description

Positioning system

The present invention relates to a positioning system, a landscape image database, a database construction device, a database construction method, a landscape image database construction program, and a positioning system for identifying the current position of the mobile body based on a landscape image photographed by an imaging means mounted on the mobile body. The present invention relates to a positioning target device.

A global positioning system (GPS) is generally known as a technique for identifying the position of a moving object. In GPS, a radio wave transmitted from a GPS satellite is received by a vehicle-mounted receiving device, and positioning is performed based on a time difference between transmission and reception.
A positioning system based on wireless technology such as GPS has a problem that positioning cannot be performed in a place where a radio wave more than the required number cannot be received. Specific examples include valleys of buildings and underpasses. In urban areas, it is easy for positioning to occur.
As a positioning technique with completely different principles that can avoid such problems, the current position of the moving object is checked by comparing a landscape image acquired by a camera mounted on the moving object with a database relating to a previously stored landscape image. Techniques for identifying are disclosed.
Patent Document 1 discloses the following positioning method. That is, from the landscape image acquired by the camera mounted on the mobile body, shape data indicating the planar shape of the road and information such as the height and color of surrounding buildings are extracted, and these are compared with a stored database. It identifies the current position of the moving object.
In Patent Document 2, the position of a feature point of a road marking is extracted from a landscape image acquired by a camera mounted on a moving body, and this is compared with a stored database to identify the current position of the moving body. A method is disclosed.
Patent Document 3 discloses a method for identifying a current position and posture by capturing a ceiling with a camera mounted on a moving object and checking the image against a stored database for a moving object operating in an indoor environment. ing.
In addition, although the purpose is not positioning, Non-Patent Document 1 detects the following obstacles as a technique for collating with a database based on feature points extracted from a landscape image, as in Patent Document 2. A technique is disclosed. That is, in this method, a landscape image acquired by a mobile body mounted camera is associated with a stored database with respect to a feature point called SIFT. An obstacle is detected by calculating a difference between road surface area images determined to correspond. Here, SIFT is an abbreviation for Scale-invariant feature transform. In this method, RANSAC (RANdom Sample Consensus) is a trial-and-error parameter estimation method as a mechanism for correctly associating a landscape image with a stored database even if erroneous feature points are mixed. Has been introduced.
Patent Document 4 discloses the following change area recognition device based on a landscape image acquired by a camera mounted on a moving body. This is done by matching the acquired landscape image against a stored database, and extracting the area that did not match after registration as a change area, that is, an area where an object that did not exist at the time of database creation exists. It is.

Japanese Patent No. 4206036 JP 2007-107043 A JP 2004-12429 A Japanese Patent No. 3966419

However, there are actually various problems when trying to determine the current position of the moving object using the related technique described above.
First, since the technique disclosed in Patent Document 3 can only be applied in an indoor environment, it cannot be applied to a moving body that functions outdoors, such as an automobile. Patent Document 2 also has a problem that it does not function unless it is an environment where road markings exist.
Patent Document 1 is considered to be a technique that can be applied regardless of location because collation is performed based on information such as a predetermined feature amount related to the appearance of a road and a building adjacent to the road. However, features such as the road width and road length (distance from the intersection to the next intersection) and the number of buildings existing on each side of the road are automatically stored in the computer. It is difficult for the current image recognition technology to extract them. Therefore, since it is necessary to manually construct the database, there is a problem that the construction cost of the database is high.
On the other hand, the method of specifying the current position of the moving body based on the position information of the feature point extracted from the landscape image as in Non-Patent Document 1 is advantageous in that it can be applied in a wide range of environments regardless of indoors and outdoors. It is. In addition, the feature point extraction can be easily realized by an existing method, so that it is effective in that the database construction cost is low.
However, such a method based on feature point matching has the following problems. In other words, after estimating the correspondence between the feature points extracted from the landscape image and the feature points stored in the database, the estimated correspondence is found for all the feature points that are randomly extracted. If it is not completely correct, the current position cannot be estimated. Here, the correct correspondence means that the feature point on the landscape image corresponding to the same part of the same object in the real world is associated with the feature point in the database.
Since it is difficult to match 100% correctly with the feature point matching process using the existing image recognition technology, the method disclosed in Non-Patent Document 1 or the like introduces a trial-and-error process called RANSAC. By this RANSAC, a set of a certain number of feature points is extracted many times, and the case where all feature points are correctly associated is included by chance.
As the ratio of feature points that are erroneously associated with each other increases, a certain number of feature point sets must be extracted more times, and the processing time required for position identification increases. There are the following problems when a vehicle or a person whose location is likely to move in an image used for constructing a landscape image database, and a plant such as a roadside tree whose shape changes as it grows. There is a high possibility that a correct correspondence relationship does not exist at the time of positioning of the positioning target moving body with respect to the feature points extracted from the region corresponding to the moving object or the like. Conventionally, however, these feature points are also handled as feature points to be collated, which increases the processing time required for position identification.
(Object of invention)
In order to solve the above-described problems, an object of the present invention is to provide a positioning system that enables a positioning process in a positioning target moving body with a smaller amount of calculation.

The landscape image database of the present invention stores a plurality of landscape image data and an image acquisition position that is a position where the landscape image data is collected, and each of the plurality of landscape image data is located in the real world. Alternatively, it includes feature amounts of feature points for things other than things whose shape is less likely to be maintained for the predetermined period or longer.
The database construction method of the present invention captures a landscape image, acquires current position information, extracts feature quantities of feature points from the landscape image, and objects whose position or shape will change in the real world from the landscape image in the future The region corresponding to is extracted.
The landscape image database construction program of the present invention is characterized by a first imaging step for capturing a landscape image, a position information acquisition step for acquiring current position information, and a feature point feature from the landscape image acquired by the first imaging means. A first feature point extracting step for extracting a quantity; a future variation area extracting step for extracting an area corresponding to an object whose position or shape will change in the real world in the future from the landscape image acquired by the first imaging means; ,
Is executed on the computer.

As described above, according to the present invention, it is possible to provide a positioning system that enables a positioning process in a positioning target moving body with a smaller amount of calculation.

It is a figure explaining this invention. It is a figure which shows an example of the description of a landscape image database. It is a figure which shows an example of the image ahead of the moving body which the 1st imaging module 101 acquired. It is a figure which shows the position of the feature point extracted from the image which the 1st imaging module 101 acquired, and there. It is a figure which shows the positional relationship of 4 points | pieces which collected the data currently recorded on the landscape image database 104, and the present position of the positioning object mobile body mounting apparatus 110. FIG. It is a figure which shows an example of the probability of the collation process with respect to four points, the positional information on a point, and the relative position (direction) information obtained by the collation process. It is a figure which shows the flowchart of the mobile body mounting apparatus for database construction. It is a figure which shows the flowchart of a positioning object mobile body mounting apparatus. It is a figure explaining the 2nd Embodiment of this invention. It is a figure which shows the flowchart of the positioning object mobile body mounting apparatus in the 2nd Embodiment of this invention. It is a figure explaining the 3rd Embodiment of this invention. It is a figure which shows the flowchart of the mobile body mounting apparatus for database construction in the 3rd Embodiment of this invention. It is a figure which shows the flowchart of the server apparatus in the 3rd Embodiment of this invention. It is a figure explaining the 4th Embodiment of this invention. It is a figure which shows the flowchart of the server apparatus in the 4th Embodiment of this invention. It is a figure which shows the flowchart of the positioning object mobile body mounting apparatus in the 4th Embodiment of this invention. It is a figure which shows the other flowchart of the server apparatus in the 4th Embodiment of this invention. It is a figure which shows the hardware constitutions of the mobile body mounting apparatus for database construction in the 4th Embodiment of this invention. It is a figure which shows the hardware constitutions of the positioning object mobile body mounting apparatus in the 4th Embodiment of this invention. It is a figure which shows the hardware constitutions of the mobile body mounting apparatus for database construction in the 4th Embodiment of this invention. It is a figure which shows the hardware constitutions of the server apparatus in the 4th Embodiment of this invention. It is a figure which shows the hardware constitutions of the server apparatus in the 4th Embodiment of this invention. It is a figure which shows the hardware constitutions of the positioning object mobile body mounting apparatus in the 4th Embodiment of this invention. It is a figure which shows the structure of the mobile body for database construction in the 5th Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(First embodiment)
A first embodiment of the present invention will be described in detail with reference to the drawings.
Referring to FIG. 1, the first embodiment of the present invention includes a database construction mobile mounting device 100 and a positioning target mobile mounting device 110. The database construction mobile is, for example, a car or robot dedicated to database construction. The positioning target moving body is, for example, a private vehicle or a business vehicle, or a robot provided with moving means such as wheels and feet.
The database construction mobile body mounting apparatus 100 includes a first imaging module 101 that captures a landscape image, and a position information acquisition module 102 that acquires the current position of the database construction mobile body.
Furthermore, the database construction mobile body mounting apparatus 100 includes a first feature point extraction module 103 that extracts feature amounts of feature points from a landscape image acquired by the first imaging module 101. Furthermore, the database construction mobile body mounting device 100 includes a landscape image database 104 that stores the feature quantities of the extracted feature points and the positional information in association with each other. The feature points to be extracted by the first feature point extraction module 103 and their feature amounts include, for example, well-known SIFT (Scale-invariant feature transform) features and SURF (Speeded Up Robust Features) features.
Further, the database construction mobile unit mounting apparatus 100 includes a future variation area extraction module 105 that extracts an area whose position or shape is likely to change in the near future from a landscape image captured by the imaging module. Areas that are likely to change in the near future are areas that record not only people and vehicles that are currently moving, but also people and vehicles that are temporary even if they are stationary and can move after a certain period of time. It is.
Furthermore, the database construction mobile unit mounting apparatus 100 includes a landscape image database construction module 106 having the following functions. That is, the landscape image database construction module 106 stores the feature quantities of the feature points extracted from areas other than those extracted by the future variation area extraction module 105 in the landscape image database in association with the current position information.
The positioning target moving body mounting apparatus 110 includes a second imaging module 111 that captures a landscape image, and a second feature point extraction module 112 that extracts feature amounts of feature points from the landscape image acquired by the second imaging module. including. Furthermore, the positioning object mobile unit mounting apparatus 110 includes an image collation / position identification module 113 and a landscape image database 114. The image collation / position identification module 113 collates the feature points extracted by the second feature point extraction module 112 with the feature point information in the landscape image database 104 to identify the current position of the positioning target mobile unit mounting apparatus 110. To do.
First, the configuration of the database construction mobile unit mounting apparatus 100 will be described in detail.
The first imaging module 101 is composed of an in-vehicle camera and the like, and captures a landscape image. For example, while the moving object for database construction moves on a predetermined road course, a landscape image in front of the vehicle is taken every moment.
The position information acquisition module 102 includes a highly accurate positioning module using RTK-GPS (Real-Time Kinetic GPS) and vehicle speed pulse information mounted on the database construction mobile body. The position information acquisition module 102 acquires position information of a point where the first imaging module 101, that is, the database construction mobile body has taken an image. The position information is expressed by two types of numerical values, for example, latitude and longitude.
The first feature point extraction module 103 extracts feature points from the landscape image acquired by the first imaging module 101, and extracts the coordinate positions and feature amounts of the feature points on the image as feature point information. As the feature points and feature amounts extracted by the first feature point extraction module 103, for example, well-known SIFT (Scale-invariant feature transform) features, SURF (Speeded Up Robust Features) features, and the like can be used.
The landscape image database 104 is a database that stores the feature point information extracted from the landscape image and the shooting position of the landscape image in association with each other.
An example of the description of the database is shown in FIG. The record number 201 is composed of latitude 203 and longitude 204 information and feature point information 205 as shooting position information 202 for a landscape image taken at a certain time. The feature point information 205 includes a set of sets of a feature point number 206 indicating the total number of feature points extracted at this time, a coordinate position 207 in the image of each feature point, and an image feature value 208 near the feature point.
For example, a SURF feature or the like may be used as the image feature amount 208. For example, the image feature amount 208 is expressed by a 64-dimensional vector value obtained by quantifying the change direction and change intensity of the pixel value in the local region. Also, as many records as the number of processed landscape images are created. In this embodiment, the landscape image database 104 is initially constructed in the database construction mobile unit mounting apparatus 100, but is then copied and placed in the positioning target mobile unit mounting apparatus 110 described later.
The future fluctuation area extraction module 105 extracts an area on the image corresponding to an object that is likely to change in the future position and shape in the real world, which is included in the image acquired by the first imaging module 101. In an outdoor environment, vehicles and pedestrians are examples of objects that are likely to change in position and shape in the future.
The future fluctuation area extraction module 105 extracts an area corresponding to a vehicle or a pedestrian by an image recognition method or the like. Specifically, it can be detected by an existing person detection method, vehicle detection method, or general object detection method.
Note that the area extracted by the future fluctuation area extraction module 105 is not an area corresponding to an object that is moving at the time when the first imaging module 101 acquires an image, but a movable object such as a person or a vehicle. Is an area corresponding to. In other words, at the time when the first imaging module 101 acquires an image, an area corresponding to a stationary person or vehicle is also extracted by this module.
An example of the area extracted by the future fluctuation area extraction module 105 will be described with reference to FIG. FIG. 3 is a diagram illustrating an example of an image ahead of a moving object acquired by the first imaging module 101, and includes a road surface area 301, a building area 302, and a parked vehicle area 303. In this case, the future change area extraction module 105 extracts the parked vehicle area 303 including the parked vehicle that may change in the future.
The landscape image database construction module 106 removes the feature points extracted from the region corresponding to the vehicle or pedestrian extracted by the future variation region extraction module 105 from the feature points extracted by the first feature point extraction module 103. The point information is processed as follows. That is, the landscape image database construction module 106 records this feature point information in the landscape image database 104 in association with the location information of the database construction mobile body acquired by the location information acquisition module 102. The feature point selection by the landscape image database construction module 106 will be described with reference to FIG. FIG. 4 is a diagram illustrating an image acquired by the first imaging module 101 and the positions of feature points extracted therefrom. The star in the figure indicates the position of the feature point. Information about the feature points 402 excluding the feature points 401 corresponding to the parked vehicle region 303 extracted by the future variation region extraction module 105 is recorded in the landscape image database 104.
Next, the configuration of the positioning target moving body mounting apparatus 110 will be described.
The second imaging module 111 is mounted on the positioning target moving body and captures a landscape image every moment.
The second feature point extraction module 112 extracts feature points from the landscape image acquired by the second imaging module 111, and extracts coordinate position information and image feature amounts on the images. The same feature point extraction algorithm as that of the first feature point extraction module 102 may be used.
The image collation / position identification module 113 collates the feature point information extracted by the second feature point extraction module 112 with the feature point information in the landscape image database 104 to identify the current position of the positioning target moving body.
Hereinafter, a specific matching method of feature points and a position identification method of the positioning target moving body will be described in detail.
First, the feature amount is extracted from the feature point group extracted by the second feature point extraction module 112 and one record in the landscape image database 104, that is, the feature point group stored in association with one point. A set of similar feature points is extracted. The feature amount of the feature point is expressed by a vector value. Therefore, specifically, for example, regarding the feature quantity of one feature point in the feature point group extracted by the second feature point extraction module 112 and the feature quantity of one feature point in the landscape image database 104, 2 Calculate the norm of the difference between two vectors. Then, it is only necessary to determine that the size is equal to or less than the threshold and to extract the corresponding feature points as a set.
After extracting all pairs of feature points having similar feature amounts, eight pairs are extracted at random. Thereafter, an estimated value of the relative positional relationship of the moving object with respect to the moving object position associated with the record in the landscape image database 104 is obtained using an algorithm called the 8-point method. In addition, the probability of the estimated relative position can be calculated using information on a set of all feature points having feature quantities similar to the estimated value.
However, in order to accurately obtain the relative positional relationship here, it is necessary that all eight sets of feature points correspond correctly. However, there is a possibility that some of the pairs of feature points may contain incorrect correspondences. Therefore, the maximum probability is calculated by changing the way of selecting the eight sets many times and calculating the relative position and the probability. Find and select the relative position that gives.
Further, the above is collation with respect to one record selected from the landscape image database 104, and collation is performed with feature points corresponding to other records, and about 1 to 3 results regarding a record with a high probability of estimation are selected. Thereafter, the current relative position with respect to the position where the data in the record is photographed is calculated according to the principle of triangulation, and the current position of the moving body is specified in consideration of the absolute position information obtained by photographing the data in the landscape image database 104.
A specific example will be described with reference to FIGS. FIG. 5 is a diagram schematically showing four points ( points 501, 502, 503, and 504) from which data recorded in the landscape image database 104 is collected and the current position 507 of the positioning target moving body. FIG. 6 shows the accuracy of the matching process between the feature point data corresponding to the points 501 to 504 and the feature point data extracted at the current time, the position information from the point 501 to the point 504, and the matching process. It is a figure which shows the value of the azimuth | direction which is the obtained relative positional relationship.
The relative position information of the base obtained by the 8-point method includes the three-dimensional relative azimuth information or absolute distance information of the moving body, and the light of the imaging module when the landscape image database 104 is created and the current positioning target moving body. This is three-dimensional rotation information of the shaft. However, FIG. 6 simply shows only relative orientation information on the horizontal plane.
In the case of FIGS. 5 and 6, for example, the determination can be made based on the relative position information extracted at the points 502 and 503 where the relatively high “probability” is extracted. 5 and 6, the angle 505 of the current position with respect to the point 502 estimated by the image collation / position identification module 113, and the optical axis of the current position with respect to the point 503 estimated by the image collation / position identification module 113. An angle 506 with respect to is shown. That is, a point (point indicated by 507 in FIG. 5) that is on an angle 505 that is 80 degrees rightward with respect to the point 502 and that is on an angle 506 that is 45 degrees rightward with respect to the point 503 is a positioning target moving body. It can be geometrically determined that the current position of the current position.
Next, the operation of the first embodiment will be described in detail with reference to the drawings. 7 and 8 are flowcharts showing the operations of the database construction mobile mounting device 100 and the positioning target mobile mounting device 110 of the present embodiment, respectively.
First, the operation of the database construction mobile unit mounting apparatus 100 will be described with reference to FIG.
First, the first imaging module 101 captures an image in front of the moving body (step S701). Further, the position information acquisition module 102 acquires the current accurate position information of the mobile body on which the database construction mobile body mounting apparatus 100 is mounted (step S702). Subsequently, the first feature point extraction module 103 extracts feature points from the image captured by the first imaging module 101 (step S703). Further, the future fluctuation area extraction module 105 extracts an area corresponding to a vehicle and a person from the image captured by the first imaging module (step S704).
Then, the landscape image database construction module 106 extracts only feature points that do not belong to the region corresponding to the vehicle or person extracted by the future variation region extraction module 105 from the feature points extracted by the first feature point extraction module 103. To do. After that, the feature point position and feature amount information relating to these feature points and the position information acquired by the position information acquisition module 102 are associated with each other and stored in the landscape image database 104 (step S705).
Steps S701 to S705 are repeated each time the first imaging module 101 acquires a new image. Note that step S702 is most preferably executed in synchronization with step S701.
Next, the operation of the positioning target moving body mounting apparatus 110 will be described with reference to the flowchart of FIG.
The second imaging module 111 captures an image in front of the positioning target moving body mounting apparatus 110 (step S801). The second feature point extraction module 112 extracts feature points from the image captured by the second imaging module 111 (step S802).
The image collation / position identification module 113 collates the feature points extracted by the second feature point extraction module 112 with the feature point information in the landscape image database 104 to identify and output the current position of the positioning target moving body. (Step S803).
The database construction mobile unit mounting apparatus 100 described above includes an ECU (Electronic Control Unit) 1801, an in-vehicle camera 1802, a hard disk 1803, and a high-precision GPS 1804 as shown in FIG. The above-described database construction mobile unit mounting apparatus 100 is described as a configuration in which the above-described module is mounted and operated in the apparatus of this configuration. The ECU 1801 controls the entire apparatus, and includes, for example, a CPU (Central Processing Unit), a RAM (Random Access Memory), a ROM (Read Only Memory), a signal processing circuit, a power supply circuit, and the like. In other words, the above-described embodiment is realized by reading out and executing a computer program capable of realizing the functions of the flowchart and the determination logic referred to in the description to the ECU 1801 except for the functions of the imaging module and the position information acquisition module. . It is also possible to configure a microcomputer by implementing the functions executed by the ECU 1801 as hardware. Furthermore, some functions may be realized by hardware, and similar functions may be realized by cooperative operation of the hardware and the software program. The hard disk 1803 is a device for storing a landscape image database, and there is no problem even if it is constituted by a storage medium other than the hard disk such as a flash memory.
The above-described positioning target mobile unit mounting apparatus 110 includes an ECU (Electronic Control Unit) 1901, an in-vehicle camera 1902, and a hard disk 1903 as shown in FIG. 19. The above-described module is mounted on the apparatus of this configuration. It has been described as a configuration that operates. The ECU 1901 controls the entire apparatus, and includes, for example, a CPU, RAM, ROM, signal processing circuit, power supply circuit, and the like. That is, the above-described embodiment is realized by reading out and executing a computer program capable of realizing the functions of the flowchart and the determination logic referred to in the description, except for the function of the second imaging module, in the ECU 1901. It is also possible to configure a microcomputer by implementing the functions executed by the ECU 1901 as hardware. Furthermore, some functions may be realized by hardware, and similar functions may be realized by cooperative operation of the hardware and the software program. A copy of the landscape image database constructed by the database construction mobile body may be stored in the hard disk 1903. Further, as another configuration example, it may be stored in a ROM in the ECU 1901 without using a hard disk.
According to the present embodiment, the landscape image database 104 is constructed by removing specific feature points from the feature points extracted by the first feature point extraction module 103. The specific feature point is a feature point belonging to a vehicle or a pedestrian that causes a reduction in the efficiency of image collation in the image collation / position identification module in the positioning target moving body mounting apparatus 110 when the future position moves. Accordingly, the image collation / position identification module 113 can correctly collate images with a smaller number of RANSAC trials.
In the above embodiment, the description has been made assuming that the image acquired by the first imaging module is an image capturing a visible light wavelength band such as a color image or a monochrome image. However, the first imaging module may be configured by an apparatus that can also acquire an image in a wavelength band other than the visible light wavelength region, such as a multispectral camera.
In this case, the future fluctuation region extraction module 105 may further extract plants and the like. Many feature points are extracted from the area of the leaves of the street tree, but the shape of the street tree changes due to wind and growth, and the position of the feature point moves, which causes noise when performing image matching .
It is not always easy to accurately determine a region corresponding to a plant such as a roadside tree from only information in the visible light wavelength region. However, since chlorophyll is known to reflect light in the near-infrared wavelength band well, using a multispectral image that includes information in the near-infrared wavelength band makes it relatively easy to identify areas corresponding to plants. Can be extracted.
Specifically, for example, a region where the reflection intensity with respect to the near-infrared wavelength band is equal to or greater than a certain threshold value may be extracted. Although fibers such as clothes often show high reflection intensity in the near-infrared wavelength band, the future fluctuation area extraction module 105 is also a detection target for pedestrians, and it is necessary to distinguish plant areas from pedestrian areas. There is no functional problem.
In the first embodiment of the present invention, a vehicle or a person who is likely to move in the future is extracted from the image used for constructing the landscape image database by the future variation area extraction module 105. Further, a region corresponding to a plant such as a roadside tree whose shape changes with growth is extracted by the future variation region extraction module 105.
After that, since the landscape image database 104 is constructed using only the information on the feature points extracted from outside these regions, the probability of selecting a pair of feature points that correspond correctly in the RANSAC process increases, and the number of RANSAC iterations is increased. Can be reduced. RANSAC is an abbreviation for RANdom Sample Consensus.
As a result, in the positioning system that identifies the current position of the positioning target moving body by comparing the position of the feature point with the database, positioning can be performed with a smaller amount of calculation.
The effect of reducing the amount of calculation will be described in detail. Assume that all the feature points extracted by the first feature point extraction module 103 are stored in the landscape image database 104, and 100 of them are associated with the feature points in the landscape image database 104. Also, 60% of them are correct.
Then, assuming that the current position can be correctly identified when all of the 8 pairs randomly extracted from the 100 pairs of feature points are correct correspondences, the probability of correctly positioning in one RANSAC trial is approximately (0. 6) ⁸ = 1.7%. Therefore, it is considered that the expected value can be measured correctly once after 60 trials.
On the other hand, when a landscape image database is constructed excluding feature points corresponding to people, vehicles, roadside trees, etc., the probability of a pair of feature points that correspond correctly increases, so this probability is assumed to be 0.8. If so, the probability of correct positioning in one trial is approximately (0.8) ⁸ = 17%. Therefore, as an expected value, positioning can be correctly performed once in six trials. That is, in the case of this example, the amount of calculation required for image collation for positioning in the positioning target moving body can be reduced to 1/10.
Further, the effect of reducing the amount of calculation is actually an effect of reducing the cost of the positioning object mobile unit mounting apparatus. This is because, by reducing the amount of calculation, the positioning object mobile unit mounting apparatus can be configured using a lower-cost embedded processor.
In the first embodiment, the future variation region extraction module 105 that extracts regions corresponding to vehicles, people, and plants is applied to the database construction mobile body mounting device 100 for the following reason. That is, it is assumed that the database construction mobile unit mounting apparatus 100 is a small number of special mobile units that can be equipped with expensive and high-performance apparatuses. This is because it is easy to tolerate price increases. In addition, in a configuration in which a landscape image database is constructed by a server device to be described later, there is no restriction on real-time processing when constructing a landscape image database. It is because it does not become.
(Second Embodiment)
A second embodiment of the present invention will be described with reference to FIG. Referring to FIG. 9, in the second embodiment, in addition to the configuration of the apparatus in the first embodiment shown in FIG. 1, the positioning target moving body mounting apparatus 910 includes the approximate position acquisition module 901. The functions of the first embodiment shown in FIG. 1 are the same except for the approximate position acquisition module 901 and the image collation / position identification module 113.
The approximate position acquisition module 901 includes an inexpensive GPS or map matching mechanism, and acquires the current approximate position of the positioning target moving body.
The image collation / position identification module 113 collates the feature point information extracted by the second feature point extraction module 112 with the landscape image database 104, and determines the current position of the moving body with higher accuracy than the approximate position. However, in the present embodiment, only records in the landscape image database 104 stored in association with positions near the approximate position acquired by the approximate position acquisition module 901 are used for collation. This makes it possible to limit the data to be collated and reduce the amount of calculation required for collation.
Next, the operation of the second embodiment will be described in detail with reference to the drawings. Since the operation of the database construction mobile unit mounting apparatus 100 is the same as that of FIG. 3, the operation of the positioning target mobile unit mounting apparatus 110 will be described in detail with reference to the flowchart of FIG.
First, the second imaging module 111 captures an image in front of the positioning target moving body (step S1001). The second feature point extraction module 112 extracts feature points from the image acquired in step S1001 (step S1002). The approximate position acquisition module 1001 acquires the current approximate position of the positioning vehicle (step S1003). The image collation / position identification module 113 collates with the feature points extracted in step S1002 using only the feature point information in the landscape image database narrowed down by the approximate position information obtained in step S1003, and Specify and output the exact current position.
According to the second embodiment, the feature point information in the landscape image database to be collated by the image collation / position identification module can be narrowed down, so that the amount of calculation required for collation can be reduced.
(Third embodiment)
A third embodiment of the present invention will be described with reference to FIG. Referring to FIG. 11, the third embodiment is different in the following points. That is, as compared with the first embodiment shown in FIG. 1, the module in the database construction mobile unit mounting apparatus 100 is divided into two, and a new database construction mobile unit mounting apparatus 1100 and a server apparatus 1120 are provided. The installation point is different. Further, a position / image data recording module 1102 and a position / image data storage unit 1103 are added as a mechanism for exchanging landscape images and position information between the database construction mobile unit mounting apparatus 1100 and the server apparatus 1120. Is different.
First, a database construction mobile body mounting apparatus 1100 according to the third embodiment will be described.
The first imaging module 101 and the position information acquisition module 102 are the same as those in the embodiment shown in FIG.
The position / image data storage unit 1103 is data in which the first imaging module 101 and the position information acquisition module 102 store image information and position information acquired at the same time in association with each other.
The position / image data recording module 1102 is a module that records the images and position information acquired by the first imaging module 100 and the position information acquisition module 101 at the same time in the position / image data storage unit 1103.
Next, the server device 1120 will be described.
The first feature point extraction module 103 extracts feature points from the landscape image recorded in the position / image data storage unit 1103, and outputs position information and feature points of feature points in the image.
The landscape image database 104 is the same as the embodiment shown in FIG. The main scenery image database 104 is once constructed on the server, but after the database construction, it is copied and installed on the positioning vehicle.
The future variation area extraction module 105 extracts an area corresponding to a vehicle, a pedestrian, or a plant whose future position or shape may change from a landscape image in the position / image data 803 referred to by the first feature point extraction module 103. To do.
The scenery image database construction module 106 records specific feature points among the feature points extracted by the first feature point extraction module 103 in association with the position information acquired by the position information acquisition module 102 in the landscape image database 104. To do. The specific feature points are other feature points excluding the feature points extracted from the regions corresponding to the vehicle, the pedestrian, and the plant extracted by the future variation region extraction module 105.
The positioning object moving body mounting apparatus 110 is the same as that of the embodiment shown in FIG.
Next, the operation of the third embodiment will be described in detail with reference to the drawings. Since the operation of the positioning object moving body mounting apparatus 110 is the same as that shown in the flowchart of FIG. 8, the description thereof is omitted. Here, the operations of the database construction mobile body mounting apparatus 1100 and the server apparatus 1120 are shown.
First, the operation of the database construction mobile unit mounting apparatus 1100 is shown in the flowchart of FIG.
The first imaging module 101 captures an image in front of the moving body (step S1201). The position information acquisition module 102 acquires the current accurate position information of the moving body in synchronization with step S1201 (step S1202). The position / image data recording module 1102 records the image and position information acquired in steps S1201 and S1202 in association with each other in the position / image data storage unit 1103 (step S1203).
Next, the operation of the server apparatus 1120 will be described with reference to the flowchart of FIG.
First, the first feature point extraction module 101 extracts feature points from the image data of the position / image data generated by the database construction mobile body mounting apparatus 1100, and extracts the position and feature amount on the image ( Step S1301). Next, the future variation area extraction module 105 extracts areas corresponding to vehicles, pedestrians, plants, and the like from the same image data as the image data referenced in step S1301 (step S1302).
Then, the landscape image database construction module 106 removes points corresponding to the region extracted in step S1302 from the feature points extracted in step S1301. A landscape image database is constructed by associating the feature amount information (position and feature amount on the image) regarding the feature point after removal with the imaging position information stored in association with the image data referred to in step S1301. (Step S1303).
The database construction mobile body mounting apparatus 1100 described above includes an image recording apparatus 2001, an in-vehicle camera 2002, a hard disk 2003, and a high-accuracy GPS 2004 as shown in FIG. 20, and is configured to mount and operate the modules described above. As explained.
The image recording apparatus 2001 controls the entire apparatus, and includes, for example, a CPU, a RAM, a ROM, a signal processing circuit, a power supply circuit, and the like. That is, in the present embodiment described above, it is realized by reading out and executing the computer program capable of realizing the functions of the flowchart and the determination logic referred to in the description to the ECU 2001, except for the functions of the imaging module and the position information acquisition module. The It is also possible to configure a microcomputer by implementing the functions executed by the ECU 2001 as hardware. Furthermore, some functions may be realized by hardware, and similar functions may be realized by cooperative operation of the hardware and the software program.
Further, the server device 1120 described above is configured by a computer 2101 as shown in FIG. 21, and has been described as a configuration in which the module described above is mounted and operated in the device of this configuration. The computer 2101 includes, for example, a CPU, RAM, ROM, signal processing circuit, power supply circuit, hard disk, and the like. That is, the above-described embodiment is realized by reading out and executing a computer program capable of realizing the function and determination logic of the flowchart referred to in the description on the computer 2101. In addition, it is possible to configure a microcomputer by implementing functions executed by the computer 2101 as hardware. Furthermore, some functions may be realized by hardware, and similar functions may be realized by cooperative operation of the hardware and the software program. The landscape image database constructed by the database construction mobile body may be exchanged with the database construction mobile body via the hard disk in the computer 2101.
Moreover, what is necessary is just to set the apparatus mounted in the vehicle for positioning to the structure similar to FIG.
According to the present embodiment, image and position information used for constructing a landscape image database is temporarily stored as position / image data, and is moved to the server device 1120 for processing. Therefore, by strengthening the calculation capability of the server apparatus 1120, there is an effect that the processing of the future variation area extraction module 105 having a relatively large calculation amount can be processed at higher speed. In particular, when the types of areas to be extracted by the future fluctuation area extraction module 105 are increased, this embodiment in which the calculation capability is easily enhanced is suitable.
(Fourth embodiment)
A fourth embodiment of the present invention will be described with reference to FIG. Referring to FIG. 14, the fourth embodiment further includes a server download module 1401 in the server device 1420 with respect to the configuration shown in FIG. Further, the positioning target mobile unit mounting apparatus 1410 has an approximate position acquisition module 901 that acquires the current approximate position, and a positioning target mobile unit download module that can communicate with the server unit 1420 to acquire a part of the landscape image database by communication. 1402 is provided.
The functions of the database construction mobile body mounting device 1400 and the server device 1420 are substantially the same as those in the example of FIG. 11, and only the server download module 1401 is different.
The server download module 1401 operates as follows when there is a request for the approximate position information and the landscape image database download from the positioning target mobile unit download module 1402 of the positioning target mobile unit mounting apparatus 1410. That is, the server download module 1401 extracts record data generated in the vicinity of the approximate position from the landscape image database stored in the server, and transmits it to the positioning target mobile unit download module 1402.
Next, the function of the positioning object moving body mounting apparatus 1410 will be described.
The functions of the second imaging module 111 and the second feature point extraction module 112 are the same as those in the embodiment of FIG.
The positioning target mobile unit download module 1402 sends the information on the approximate position acquired by the approximate position acquisition module 601 and the request message for landscape image data to the server download module 1401 of the server device 1420. Thereafter, the positioning object moving body download module 1402 receives the corresponding landscape image data from the server download module 1401.
The image collation / position identification module 113 collates the feature point information extracted by the second feature point extraction module 112 with the landscape image data received from the server download module 1401 to determine the accurate current position of the moving object.
Next, the operation of the present embodiment will be described. The operation of the database construction mobile unit mounting apparatus is the same as the flowchart shown in FIG.
The operation of the server device 1420 will be described with reference to the flowchart of FIG. Steps S1501 to S1503 are the same as steps S1201 to S1203.
The server download module 1401 operates as follows only when a landscape image database is requested from the positioning target mobile unit download module 1402 of the positioning target mobile unit mounting apparatus 1410. That is, the server download module 1401 extracts and transmits corresponding landscape image data (step S1504).
Next, the operation of the positioning target moving body mounting apparatus 1410 will be described with reference to the flowchart of FIG.
First, the second imaging module 111 captures a landscape image (step S1601). The second feature point extraction module 112 outputs the feature amount of the feature point from the image captured in step S1601 (step S1602). The approximate position acquisition module 901 acquires the approximate position of the positioning target moving body mounting apparatus 1410 (step S1603).
The positioning target mobile unit download module 1402 communicates with the server download module 1401 of the server device 1420 and receives landscape image data corresponding to the approximate position (step S1604).
Finally, the image collation / position identification module 113 collates the landscape image data received by the positioning target mobile unit download module 1402 with the feature amounts of the feature points extracted by the second feature point extraction module 112. Thereafter, the image collation / position identification module 113 determines the current position of the moving body (step S1605).
The server device 1420 described above is configured by a computer 2201 and a communication device 2202 as shown in FIG. 22, and has been described as a configuration in which the module described above is mounted and operated in the device of this configuration. The computer 2202 includes, for example, a CPU, RAM, ROM, power supply circuit, hard disk, and the like. That is, the above-described embodiment is realized by reading out and executing a computer program capable of realizing the functions and determination logic of the flowchart referred to in the description, excluding the function of the server download module 1401. In addition, it is possible to configure a microcomputer by implementing functions executed by the computer 2201 as hardware. Furthermore, some functions may be realized by hardware, and similar functions may be realized by cooperative operation of the hardware and the software program. The communication device 2202 is hardware for wireless LAN (Local Area Network) or mobile phone data communication that executes the function of the server download module 1401.
Further, the above-described positioning target moving body mounting apparatus 1410 includes an ECU (Electronic Control Unit) 2301, a camera 2302, a hard disk 2303, a GPS 2304, and a communication device 2305 as shown in FIG. In the apparatus having this configuration, the module described above is mounted and operated. The ECU 2301 controls the entire apparatus, and includes, for example, a CPU, RAM, ROM, signal processing circuit, power supply circuit, and the like. That is, the above-described embodiment is realized by reading out and executing a computer program capable of realizing the functions and determination logic of the flowchart referred to in the description to the ECU 2301. It is also possible to configure a microcomputer by implementing the functions executed by the ECU 2301 as hardware. Furthermore, some functions may be realized by hardware, and similar functions may be realized by cooperative operation of the hardware and the software program.
The landscape image database constructed by the database construction mobile body may be stored in the hard disk 2303 as a copy. As another configuration example, it may be stored in a ROM in the ECU 2301 without using a hard disk. The communication device 2305 is hardware for wireless LAN or cellular phone data communication that executes the function of the server download module 1401.
According to the present embodiment, since it is not necessary to hold the landscape image database in the positioning target mobile unit mounting apparatus 1410, the capacity of the magnetic disk mounted on the positioning target mobile unit mounting unit 1410 can be reduced. Furthermore, since the landscape image database is stored in the server device 1420, there is an advantage that the landscape image database can be easily updated. When the landscape image database is held in the positioning target mobile unit mounting apparatus 1410, it is necessary to update the landscape image database stored in all the positioning target mobile unit mounting apparatuses 1410.
In the flowchart of the server device 1420, an example of the operation of executing steps S1501 to S1504 as a series of processes has been described. However, the construction of the landscape image database and the transmission of the landscape image data may not be performed at the same time. . That is, according to the user's operation mode designation, as shown in FIG. 17, the operation may be performed to execute either the landscape image database construction mode or the landscape image data transmission mode.
(Fifth embodiment)
Next, a fifth embodiment for carrying out the present invention will be described.
FIG. 24 shows a landscape image database according to the fifth embodiment of the present invention.
The landscape image database 2401 of this embodiment stores a plurality of landscape image data and image acquisition positions that are positions where the landscape image data is collected in association with each other.
Furthermore, the landscape image database 2401 of the present embodiment is characterized in that each of the plurality of landscape image data is a feature point for an object other than an object whose position or shape in the real world is less likely to be maintained as it is for a predetermined period or longer. Includes features.
In the fifth embodiment described above, the following effects can be obtained. In other words, a landscape image of a positioning system that can perform positioning processing on a positioning target moving body with a smaller amount of calculation by identifying feature points that are likely to be difficult to correctly associate and not storing them in the landscape image database in advance. A database can be provided.
In each embodiment described so far, a dedicated device is assumed, but the following may be used. That is, for example, a personal computer device that performs various data processing is loaded with a board or a card that performs processing corresponding to this example, and each processing is executed on the computer device side. In this way, a configuration may be adopted in which software for executing the processing is installed in a personal computer device and executed.
The program installed in the data processing device such as the personal computer device may be distributed via various recording (storage) media such as an optical disk and a memory card, or distributed via communication means such as the Internet. Also good.
In addition, each of the above embodiments can be combined with other embodiments.
While the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.
This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2010-226844 for which it applied on October 6, 2010, and takes in those the indications of all here.
Some or all of the embodiments described above can be described as in the following supplementary notes, but are not limited thereto.
(Appendix 1)
A plurality of landscape image data and an image acquisition position that is a position where the landscape image data is collected are stored in association with each other,
Each of the plurality of landscape image data includes a feature value of a feature point with respect to an object other than an object whose position or shape in the real world is unlikely to be maintained as it is for a predetermined period or more. Database.
(Appendix 2)
2. The landscape image database according to appendix 1, wherein an object that is less likely to be maintained in a current position or shape in the real world for a predetermined period or longer is a vehicle.
(Appendix 3)
The landscape image database according to appendix 1 or appendix 2, wherein the thing that is unlikely to be maintained as it is for a predetermined period or longer in the real world is a person.
(Appendix 4)
A landscape image database described in any one of supplementary notes 1 to 3;
A first imaging means for capturing a landscape image;
Position information acquisition means for acquiring current position information;
First feature point extracting means for extracting feature amounts of feature points from the landscape image acquired by the first imaging means;
A future fluctuation area extracting means for extracting an area corresponding to an object whose position or shape in the real world is less likely to be maintained as it is for a predetermined period or more from the landscape image acquired by the first imaging means;
The feature image of the feature point obtained by removing the feature point extracted from the area extracted by the future variation area extraction unit from the feature point acquired by the first imaging unit is associated with the current position information and the landscape image. Means for constructing a landscape image database stored in the database;
A database construction device comprising:
(Appendix 5)
The first imaging unit captures not only an image in the visible light wavelength band but also an image in a wavelength band other than the visible light region, and the future variation area extracting unit captures the landscape image acquired by the first imaging unit. The database construction device according to appendix 4, wherein a region corresponding to a plant is extracted from the inside.
(Appendix 6)
A landscape image database described in any one of supplementary notes 1 to 3;
A second image pickup means for picking up a landscape image; a second feature point extraction means for extracting feature amounts of feature points from the landscape image acquired by the second image pickup means; and the second feature point extraction means. Image collation / position identification means for collating the feature amount of the extracted feature point with the landscape image database to identify the current position of the positioning target moving body;
A positioning target device comprising:
(Appendix 7)
The positioning target device further includes a rough position acquisition unit that acquires a current rough position of the positioning target device, and the image collation / position identification unit includes, among landscape image data stored in the landscape image database, The positioning target device according to appendix 6, wherein the positioning target device performs image matching using only landscape image data generated near the approximate position.
(Appendix 8)
The database construction device described in any one of the supplementary notes 1 to 3;
The positioning target device described in Appendix 6 or Appendix 7,
And a landscape image database of the positioning target device has the same content as a landscape image database constructed by the database construction device.
(Appendix 9)
When a position information and landscape image data transfer request is received from the positioning object mobile body, the landscape image database in the server device that transmits the landscape image data in the landscape image database associated with the position information in the vicinity of this position information Including a server device having download means,
The positioning target moving body transmits the position information and the landscape image data transfer request to the rough position acquisition means for acquiring a rough position of the positioning target moving body and the scenery image database download means in the server device, and the outline is acquired. A positioning object moving body scenery image database download means for acquiring scenery image data in the scenery image database associated with position information near the position;
The image collation / position identification unit identifies the current position by collating the landscape image data acquired by the landscape image database download unit with the feature amount of the feature point extracted by the second feature point extraction unit. The positioning system according to appendix 8, characterized by:
(Appendix 10)
Take a landscape image,
Get current location information
Extracting feature amounts of feature points from the landscape image;
An area corresponding to an object whose position or shape changes in the real world in the future is extracted from the landscape image.
A database construction method characterized by that.
(Appendix 11)
Not only images in the visible light wavelength band, but also images in the wavelength band other than the visible light range,
Extracting an area corresponding to a plant from the landscape image;
The database construction method according to supplementary note 10, characterized by:
(Appendix 12)
A first imaging step for capturing a landscape image;
A location information acquisition step for acquiring current location information;
A first feature point extracting step of extracting feature amounts of feature points from the landscape image acquired by the first imaging means;
A future change area extracting step of extracting an area corresponding to an object whose position or shape will change in the future in the real world from the landscape image acquired by the first imaging means;
A landscape image database construction program that causes a computer comprising the landscape image database described in any one of supplementary notes 1 to 3 to be executed.

The present invention relates to a positioning system, a landscape image database, a database construction device, a database construction method, and a landscape image database construction program for identifying the current position of the mobile body based on a landscape image taken by an imaging means mounted on the mobile body. In addition, the present invention relates to a positioning target device and has industrial applicability.

DESCRIPTION OF SYMBOLS 100 Mobile construction apparatus for database construction 101 1st imaging module 102 Position information acquisition module 103 First feature point extraction module 104 Landscape image database 105 Future fluctuation area extraction module 106 Landscape image database construction module 110 Positioning object mobile equipment installation apparatus 111 Second imaging module 112 Second feature point extraction module 113 Image collation / position identification module 114 Landscape image database 201 Record number 202 Shooting position information 203 Latitude 204 Longitude 205 Feature point information 206 Number of feature points 207 Coordinate position 208 Image feature amount 301 road surface area 302 building area 303 parked vehicle area 401 feature points 402 corresponding to the parked vehicle area 303 feature points 501 and 5 excluding the feature point 401 2, 503, 504 Four points where data was collected 505 Angle of the current position relative to the point 502 with respect to the optical axis estimated by the image verification / position identification module 506 Current position relative to the point 503 estimated by the image verification / position identification module Angle 507 relative to the optical axis 507 Current position of the positioning target moving body 901 Approximate position acquisition module 910 Positioning target moving body mounting apparatus 1102 Position / image data recording module 1103 Position / image data storage means 1120 Server apparatus 1401 Server download module 1402 Positioning target movement Body download module 1420 server device 1801 ECU
1802, 2002 Camera 1803, 2003 Hard disk 1804, 2004 High precision GPS
1901,301 ECU
1902, 2302 Camera 1903, 2303 Hard disk 2001 Image recording device 2101, 2201 Computer 2202 Communication device 2304 GPS
2305 Communication equipment

Claims (12)

1. A plurality of landscape image data and an image acquisition position that is a position where the landscape image data is collected are stored in association with each other,
   Each of the plurality of landscape image data includes a feature value of a feature point with respect to an object other than an object whose position or shape in the real world is unlikely to be maintained as it is for a predetermined period or more. Database.
2. 2. The landscape image database according to claim 1, wherein a thing having a low possibility that the position or shape in the real world is maintained as it is for a predetermined period or longer is a vehicle.
3. 3. The landscape image database according to claim 1 or 2, wherein a thing having a low possibility that the position or shape in the real world is maintained as it is for a predetermined period or longer is a person.
4. A landscape image database according to any one of claims 1 to 3;
   A first imaging means for capturing a landscape image;
   Position information acquisition means for acquiring current position information;
   First feature point extracting means for extracting feature amounts of feature points from the landscape image acquired by the first imaging means;
   A future fluctuation area extracting means for extracting an area corresponding to an object whose position or shape in the real world is not likely to be maintained as it is for a predetermined period or more from a landscape image acquired by the first imaging means; The feature amount of the feature point excluding the feature point extracted from the area extracted by the future variation area extracting unit from the feature point acquired by the imaging unit is stored in the landscape image database in association with the current position information A landscape image database construction means,
   A database construction device comprising:
5. The first imaging unit captures not only an image in the visible light wavelength band but also an image in a wavelength band other than the visible light region, and the future variation area extracting unit captures the landscape image acquired by the first imaging unit. The database construction device according to claim 4, wherein a region corresponding to a plant is extracted from the inside.
6. A landscape image database according to any one of claims 1 to 3;
   A second image pickup means for picking up a landscape image; a second feature point extraction means for extracting feature amounts of feature points from the landscape image acquired by the second image pickup means; and the second feature point extraction means. Image collation / position identification means for collating the feature amount of the extracted feature point with the landscape image database to identify the current position of the positioning target moving body;
   A positioning target device comprising:
7. The positioning target device further includes a rough position acquisition unit that acquires a current rough position of the positioning target device, and the image collation / position identification unit includes, among landscape image data stored in the landscape image database, 7. The positioning target apparatus according to claim 6, wherein the positioning target apparatus performs image collation only using landscape image data generated near the approximate position.
8. A database construction device according to any one of claims 1 to 3;
   A positioning target device according to claim 6 or 7, and
   And a landscape image database of the positioning target device has the same content as a landscape image database constructed by the database construction device.
9. When a position information and landscape image data transfer request is received from the positioning object mobile body, the landscape image database in the server device that transmits the landscape image data in the landscape image database associated with the position information in the vicinity of this position information Including a server device having download means,
   The positioning target moving body transmits the position information and the landscape image data transfer request to the rough position acquisition means for acquiring a rough position of the positioning target moving body and the scenery image database download means in the server device, and the outline is acquired. A positioning object moving body scenery image database download means for acquiring scenery image data in the scenery image database associated with position information near the position;
   The image collation / position identification unit identifies the current position by collating the landscape image data acquired by the landscape image database download unit with the feature amount of the feature point extracted by the second feature point extraction unit. The positioning system according to claim 8.
10. Take a landscape image,
    Get current location information
    Extracting feature amounts of feature points from the landscape image;
    An area corresponding to an object whose position or shape changes in the real world in the future is extracted from the landscape image.
    A database construction method characterized by that.
11. Not only images in the visible light wavelength band, but also images in the wavelength band other than the visible light range,
    Extracting an area corresponding to a plant from the landscape image;
    The database construction method according to claim 10.
12. A first imaging step for capturing a landscape image;
    A location information acquisition step for acquiring current location information;
    A first feature point extracting step of extracting feature amounts of feature points from the landscape image acquired by the first imaging means;
    A future change area extracting step of extracting an area corresponding to an object whose position or shape will change in the future in the real world from the landscape image acquired by the first imaging means;
    The program recording medium which recorded the landscape image database construction program characterized by making a computer perform.

Images